How to Build a High‑Availability Microservices System on Kubernetes – A Complete Guide

Microservice Design

The system follows Martin Fowler’s microservice principles of high cohesion and low coupling. A simple three‑service chain a.demo.com → b.demo.com → c.demo.com illustrates request flow. To achieve reliability each service runs multiple instances and registers with a Eureka service registry, which provides dynamic discovery and load balancing across instances.

Observability is built with Prometheus and Grafana for metrics (CPU, memory, HTTP error counts, JVM thread count, GC time, etc.). Application logs are streamed to Kafka to avoid container‑local I/O, and distributed tracing is enabled via Zipkin, which propagates a unique trace ID across service calls.

Flow‑control mechanisms—circuit breaking, rate limiting, and degradation—are implemented with Alibaba Sentinel, which integrates seamlessly with Spring Cloud.

Implementation Details

The front‑end is a static HTML page that issues an AJAX request to an Nginx‑proxied endpoint. Nginx forwards the request to a.demo.com, which in turn calls b.demo.com, and finally c.demo.com. The back‑end consists of three Maven modules (a, b, c) that share a common utility library and a Eureka server for registration.

Each service exposes a health‑check endpoint /hs so Kubernetes can perform liveness and readiness probes. Configuration files are written in YAML, allowing seamless switching between local execution and container deployment.

Kubernetes Deployment

A three‑node HA Kubernetes cluster can be provisioned with the open‑source installer K8seasy . The installer bundles required container images and supports multiple Kubernetes versions.

Generate cluster keys: sudo ./installer --genkey -hostlist=192.168.2.1 Create the cluster (example uses Kubernetes v1.18.2):

sudo ./installer -kubernetestarfile kubernetes-server-linux-amd64v1.18.2.tar.gz -masterip 192.168.2.50

Verify installation via the Kubernetes dashboard (port 10000) and access Alertmanager (8080) and Grafana (8081, default credentials admin/admin).

Docker images for the services are hosted on Docker Hub. Applying the provided YAML manifests creates Deployments, Services, and a three‑replica Eureka StatefulSet, enabling automatic start‑up and HA.

High‑Availability Verification

Verification steps include:

Open the front‑end at www.demo.com and click the request button; response times from each instance are displayed.

Run a simple script that issues a request every three seconds to simulate continuous traffic.

Inspect the Zipkin UI ( zipkin.demo.com) to view end‑to‑end trace details for each request.

Use the Sentinel console to monitor real‑time metrics, rate‑limiting counters, and circuit‑breaker status.

View Grafana dashboards for JVM heap usage, GC pauses, error counts, and overall service health.

These checks demonstrate that the failure of any single instance does not affect overall system functionality, confirming the high‑availability design.

Repository

All source code, Dockerfiles, and Kubernetes YAML manifests are available at:

https://github.com/xiaojiaqi/deploy-microservices-to-a-Kubernetes-cluster